Investigating Users' Requirements

Origin of the Visible Human Project

A key objective of the Next Generation Internet/ Visible Human (NGI/VH) Project at University of Michigan is to facilitate the use of medical images from the Visible Human data set in anatomy learning. Advances in networking technologies have enhanced the ability of users to retrieve, view, and manipulate digital medical images to satisfy needs in clinical diagnosis, research, and education.

Foreseeing the need for a large digital image library of the entire human male and female bodies, the National Library of Medicine (NLM) began the Visible Human Project in 1986.1 2 Since 1996, the Visible Human data sets, with more than 13,000 computerized tomography, magnetic resonance, and cross-sectional photography images showing the normal female and male anatomic structures, have been available from NLM to licensed users via the Internet (http://www.nlm.nih.gov/research/visible/visible_human.html). The NGI/VH Project at the University of Michigan is one of the projects currently funded to develop an instructional system to support simultaneous use of Visible Human data set resources for anatomy navigation via personal computers.

The development of the NLM Visible Human data sets inspired further research in medical imaging for health care and education. However, difficulties with content design and information delivery have been encountered in the use of raw images from the data sets—namely, a lack of accurate labels and appropriate instructions for some images, which undermines their usability and instructional effectiveness. Researchers have asserted that difficulty in using the original Visible Male data set in educational programs is due to lack of labeling of the data.18 In addition to applying labels, researchers are challenged to design a user interface that enables efficient retrieval of digital images. Interfaces invented by previous researchers vary from a form-based retrieval design4 and interactive three-dimensional atlases3 5 to an ideal surgical simulation providing haptic, or tactile, feedback to the user.1

Although the work of the University of Michigan Visible Human Project continues to build on the previous studies, the project researchers were also interested in integrating the many modalities of user interaction in Visible Human information retrieval with identified learner needs. Taking into account the validity in anatomy teaching and learning situations, the intent was to provide diverse users with the option to use distinct learning modules. A module is defined in this research as learning activities consisting of individual units of study, which correspond to key topics in gross anatomy course work and professional training.

Theoretic Framework

Design of Visible Human software suitable for the educational goals of diverse user groups was the research goal of this study. The intended users of the Visible Human software include health sciences students and faculty in nursing, medicine, dentistry, and kinesiology. One assumption that guided the study is that a strong correlation and interaction exist between the content of particular modules and the specific context for which anatomy is learned.

The importance of and need for contextual analysis of technology-based learning environments were shown in previous research on educational technology. Jacobson and Spiro6 suggested that, to promote use of technologies, researchers should study the conceptual characteristics of the knowledge area and also the stage of learning (novice or expert) of the users. Tergan7 criticized hypertext and hypermedia designed for education. He asserted that inadequate match of design to the characteristics of the subject matter and omission of cognitive demands for processing the knowledge resulted in failures of hypermedia used in learning.

In this study, the contextual analysis focused on the collection of learning-context and pedagogic-content information.8 The learning-context information described the teaching setting and teaching goals. The pedagogic-content information addressed difficulties in and processes for learning anatomy. These types of information are particularly important to this research team, which consisted of faculty and graduate students from diverse disciplines, such as nursing, education, medicine, kinesiology, and dentistry, as well as software designers and programmers.

Data were collected from both faculty and students. The need to understand the difficulties students experienced, their learning habits, and their use of study resources in learning anatomy was emphasized. Previous research suggested that the way novice students view instructional materials may differ from the way experts view them.8 Studying how a novice overcomes learning difficulties and achieves understanding provides essential information for instructional strategies.

This study served as the initial step of a formative evaluation. As opposed to summative evaluation, which focuses on effectiveness of the end product, formative evaluation uses early and frequent feedback from potential students and faculty. The evaluation emphasized measurement of users' requirements and perception. The intent was to have user data and information about anatomy education drive iterative improvement9 of the design and thus create a learning environment that is efficient and effective.

The software design gave rise to a question: After initial input from potential users has been acquired, how can these data be converted into applicable guidelines for learning-module design? Many excellent sources of information on user interface design exist.10 11 12 However, there are no fixed methods for transforming information and demands gathered about users into an effective design.

The process of transforming users' requirements into concrete designs has been a black box. This study begins to fill an important void by describing how users' requirements, associated with users' learning experiences, were systematically collected and analyzed and then transformed into guidelines informing the iterative design of multiple learning modules. This work was guided by the belief that a good instructional design should provide access to functions and features in ways that reflect how users think about the tasks that a potential application would support.12 13

Methods

Participants

To gain a comprehensive view of users' requirements, at least one focus group session was conducted with each group of target users—students and faculty in schools of nursing, medicine, dentistry, kinesiology, and surgery. In two of the health sciences schools, dentistry and medicine, multiple groups of students at different learning stages were recruited. Faculty who teach anatomy or supervise health sciences students were contacted by the researchers and invited to participate. Student participants were recruited by faculty through e-mail or in-class invitation. Participation was voluntary and anonymous. Institutional Review Board (IRB) approval was received before data collection began. Following IRB guidelines, consent to participate in the sessions was received from each student and faculty member prior to the focus group interactions.

Procedure

The nominal group technique14 was implemented by asking each participant to write short statement answers to five or six open-ended questions at the beginning of the focus group session. Each session lasted no more than 1 hour. After the 5- to 10-minute warm-up time, when participants wrote answers to the open-ended questions, each participant had a chance to recall from memory his or her previous experiences in learning anatomy. This activity ensured that individual group members each had the opportunity to contribute their own thoughts, and helped prevent the omission of important ideas because of the flow of discussion or the dominance of a particular group member.

Participants' written responses to this research instrument complemented information gathered during discussion, to provide a more comprehensive record of their ideas. When the discussion started, participants were asked to describe, in a round-robin fashion, their ideas related to those questions. At least two researchers were assigned as leaders of each focus group session. Leaders used follow-up questions and inquired about details, according to the flow of discussion and the time constraints.

Researchers took handwritten notes during the focus group sessions. Because of the demanding focus group schedule and the time constraints, notes and recollections rather than transcriptions of tapes were used as data sources. Focus groups were conducted from February to December 2000. Table 1 shows the results of focus group users' requirement ratings. Attributes among groups vary by discipline and anatomy study experience.

Data Analysis

Ideas (in the form of short statements) that students or instructors contributed about their difficulties in learning or teaching anatomy, and their visions of future applications were combined to develop a requirement list. One team member created a matrix by pooling users' requirements from all focus groups. The row headings of this matrix listed the users' requirements and the column headings the target users represented by the focus groups.

The purpose of the data analysis was to help researchers understand overall users' requirements and special needs of each target group. Each of the 11 researchers from the University of Michigan NGI/VH user interface team participated in at least two focus group sessions; they then met as a group to share and discuss information gathered in the different sessions.

The purpose of the discussion was to help members of the user interface team develop a comprehensive understanding of the anatomy teaching and anatomy learning issues expressed in the focus groups. Each researcher was asked to rate the users' requirements by assigning a number from 1 (“not important at all”) to 6 (“must include in the design”) to each. From these data, a summary matrix containing the mean rating from all evaluators for each cell, the mean rating for each row, and the mean for each column was prepared. The rows (users' requirements) data were sorted by row mean, from most to least important. Similarly, the column (target groups) data were sorted by column mean (Figure 1).

Further examination of these rating data, using cluster analysis and multidimensional scaling ,15 was performed to find target groups who received similar scores in all requirements. For multidimensional scaling, the cell means shown in Figure 1 (excluding row means and column means) were used as data to calculate the Euclidean distances between all user groups. The scaling algorithm produced two optimized, statistically generated dimensions. Two kinds of hierarchic algorithms—the complete linkage and Ward's method—were used to compare the similarity of all requirements across different users and to group comparable requirements into clusters. Figure 2 shows a flowchart of the data analysis procedure, including the three different statistical analyses that were conducted—sorting, multidimensional scaling, and cluster analysis.

Results

The results of the multidimensional scaling are shown in Figure 3. Different user groups are distributed into four quadrants, showing that users' requirements for those who share similar experience and training objectives in their health care education fall into the same quadrants of the coordinate system.

The two optimized dimensions were interpreted by comparing the attributes of the groups on the opposite ends along each dimension. All the groups above the horizontal axis have a more holistic or systemic view of anatomy, whereas those groups below have a more narrow or regional locus of interest. Thus, the vertical dimension was conceptualized as anatomic scope or breadth of focus, while the horizontal dimension was interpreted as level of professional development, novice vs. expert (right vs. left). Along the vertical axis, undergraduate kinesiology students, sophomore nursing students, first-year dental students, first-year medical students, second-year dental students, and second-year medical students are on the side opposite the fourth-year medical students, surgical residents, and all the faculty groups. The latter received advanced training in dissection and achieved greater expertise in anatomy than did the other groups.

From these data, it was inferred that the four quadrants could be labeled holistic-expert (upper left), holistic-novice (upper right), narrow-expert (lower left), and narrow-novice (lower right). These indicate at least four clusters of users that the design should accommodate. However, the multidimensional scaling analysis did not provide sufficient information to enable description of how the distinct users in the same natural group differ (e.g., first-year dental students and first-year medical students, or kinesiology faculty and nursing faculty). For this to occur, further qualitative analysis of the learning content and teaching practices must take place.

Results from the hierarchic algorithm show clusters of similar requirements (Figure 4). Each cluster was composed of numerous related functions and features that should be integrated simultaneously into the learning environment. To make the results in the summary matrix more informative for the interface design, clusters were categorized by level of demand and correlated with characteristics of the users who requested them. The three major categories were defined as general features and functions, domain-specific applications, and users' tools. Different requirement clusters and the design guidelines derived from them are shown in Table 2.

Discussion

The primary goal of the NGI/VH user interface design is to provide multiple instructional modules addressing the educational needs of different users who will be accessing the Visible Human data sets. Findings from the multidimensional scaling, cluster analysis, and requirement sorting resulted in the creation of guidelines for both interaction and content design.

Interface Design

At least two sets of functions and features should be used in the anatomy learning modules. One set would be core components and be used as the base design and framework. These components would be common across groups. Their content was identified in the focus groups, by faculty and students across the different disciplines, as important requirements for learning anatomy. In addition, natural groups of users (Figure 3) share similar but different demands in learning anatomy. The second set of functions and features would not be universal across all groups but would fulfill the needs of self-selected user groups, which parallel their specialties.

To design learning modules with both unified core components and user-specific applications, the program should be flexible, allowing for dynamic insertion of different learning tools for different users. One possible way to achieve this goal is to provide an interface that users can employ to select and enable suitable learning materials.

To turn the user interface into a self-directed learning environment, it is important to provide tools that enable students to gain easy access to required information and resources. These tools should include directions for navigation and concept mapping. During focus groups, students suggested that both the hierarchic structure of information (from broad concepts to specific details) and the mapping of concepts were important. Providing the overall organizational structure of the program to students will allow them to move among different presentations of content information without necessarily following a pre-defined navigation path through an instructional module.

Content Design

Visualization

A design theme that emerged during analysis was three-dimensional visualization and relationship reconstruction. The focus group data, like findings from previous studies,16 17 showed that students have difficulty mapping two-dimensional images, like those found in textbooks, to three-dimensional structures and live patients. Therefore, a feature that displays two- and three-dimensional images simultaneously on the screen is needed. This would allow students to compare the two different kinds of information on the same human structure. The relationship between the images should be shown explicitly by reference points. In addition, the three-dimensional models should correctly illustrate the spatial relationship of different structures (such as organs) in the same region or system.

To provide students with a virtual learning aid to facilitate the visualization process, software should provide a mechanism for representing discrete anatomic structures and be able to store information about the relationships between these structures. Hyperlinked text and images is one feature that was suggested to support whole–part relationships.

Information Navigation

In learning anatomy, students may acquire knowledge through problem-based learning and exploration of structure-function relationships. Connection of anatomy knowledge to information in other classes and to clinical cases is one aspect of information navigation.

During the focus groups, students indicated that they would like to know how anatomy relates to the content of other courses (e.g., physiology and pathology). Even though the current design of this program emphasizes its use in anatomy classes, information about related concepts taught in other classes and in clinical settings could be integrated, which would help students conceptualize health care knowledge as a whole.

Specific Applications

Certain specific applications, visualization requirements, and tools are important for selected groups of learners. Medical and surgical students, for example, spend a great deal of time in the anatomy laboratory exploring the human body and finding anatomic structures in it. Many students expressed a desire to have simulations incorporated into the program, which would allow them to practice dissection before actually performing the procedure.

Other domain-specific applications required by users were the ability to follow structures, visualization of blood supply or innervation, and dynamic visualization. A software component could be included that allows students to see from “hither to yon” if structures are connected.

Learning Aids

Because of the large volume of information presented in gross anatomy courses, many students expressed the need for some guidance function or feature to help them distinguish the most important information to learn (using some form of highlighting, for example). One suggestion was to include mnemonic aids in the training module, like those used by students in “traditional” anatomy courses.

Another requested special feature was communication tools that could be used by study group members. Because regular e-mail may be restricted by limited functions, students expressed a desire to have some other communication and collaboration tools that could facilitate information exchange.

Because learning aids may be more important to some learners than others, it may be crucial to give users the ability to display or suppress certain features in the learning environment. For example, when students become more familiar with the pronunciation of medical terminology, they should be able to turn off this function by simply checking a preference box that disables the pronunciation feature.

Instructor's Tools

Instructors requested a separate interface for the selection of anatomy content, content display, and functionality control. Unlike other, commercial electronic resources for learning anatomy, this research strives to give instructors the ability to integrate their teaching materials. Instructors from different health care disciplines have different approaches to presenting and organizing course material. With this in mind, the software should attempt to modularize the information segments as much as possible. This would allow instructors to “snap together” learning modules into a format with which they are familiar and comfortable. This component will enable instructors to choose the appropriate level of material and to guide the learning process of their students.

Conclusion

Both the research process and the results used to specify the essential users' requirements for anatomy learning tools are presented. Employing ideas and feedback from users is a powerful tool for designing and polishing the NGI/VH user interface system. Although data gathering and information analysis are important steps in the design of the user interface, they do not, in themselves, guarantee the successful design of learning environment software.

Data generated from research team members' matrixes of users' requirement ratings should be interpreted with caution, because the researcher ratings may not fully reflect the direct ideas expressed in the focus groups. On one hand, a researcher may feel more confident rating the groups they directly participated in than those they did not. On the other hand, combined analyses of matrixes from multiple researchers can reduce the bias in favor of the perspectives of one or two researchers. Also, in reality, while it is difficult to have all the researchers participate in all the focus groups, this is still one of the best ways to collect information from numerous target groups.

The learning environment design described here will continue to evolve throughout the software development process, enabling the research team to link users' needs with the ability to construct an appropriate and customized instructional design. Only by the art and science of design can a solid user interface serving the desired educational purpose be generated. Information presented here has been used to generate the initial design of the user interface and content structure. The development of a system prototype has begun. Future papers will enable the reader to trace how user input has been used to modify and improve the final design, from pilot study through prototype.